bugfix

Author: mathias-von-ottenbreit

cpp/APLRRegressor.h

@@ -1371,6 +1371,9 @@ void APLRRegressor::determine_interactions_to_consider(const std::vector<size_t>
                 if (model_term_without_given_terms_can_be_a_given_term)
                     model_term_with_added_given_term.given_terms.push_back(model_term_without_given_terms);
                 add_necessary_given_terms_to_interaction(interaction, model_term_with_added_given_term);
+                bool interaction_only_uses_one_base_term{interaction.term_uses_just_these_predictors({interaction.base_term})};
+                if (interaction_only_uses_one_base_term)
+                    continue;
                 if (interaction_constraints_provided)
                 {
                     bool interaction_violates_constraints{true};
@@ -2457,7 +2460,7 @@ std::map<double, double> APLRRegressor::get_main_effect_shape(size_t predictor_i
         return main_effect_shape;
 
     std::vector<double> split_points;
-    size_t max_potential_split_points{relevant_term_indexes.size() * 3 + 2};
+    size_t max_potential_split_points{(relevant_term_indexes.size() * 3 + 2) * 3};
     split_points.reserve(max_potential_split_points);
     for (auto &relevant_term_index : relevant_term_indexes)
     {
@@ -2478,6 +2481,22 @@ std::map<double, double> APLRRegressor::get_main_effect_shape(size_t predictor_i
     split_points.push_back(min_predictor_values_in_training[predictor_index]);
     split_points.push_back(max_predictor_values_in_training[predictor_index]);
     split_points = remove_duplicate_elements_from_vector(split_points);
+
+    VectorXd split_point_increments{VectorXd(split_points.size() - 1)};
+    for (Eigen::Index i = 0; i < split_point_increments.size(); ++i)
+    {
+        split_point_increments[i] = split_points[i + 1] - split_points[i];
+    }
+    double minimum_split_point_increment{split_point_increments.minCoeff()};
+    double increment_around_split_points{minimum_split_point_increment / DIVISOR_IN_GET_MAIN_EFFECT_SHAPE_FUNCTION};
+
+    size_t num_split_points_before_small_increments{split_points.size()};
+    for (size_t i = 0; i < num_split_points_before_small_increments; ++i)
+    {
+        split_points.push_back(split_points[i] - increment_around_split_points);
+        split_points.push_back(split_points[i] + increment_around_split_points);
+    }
+    split_points = remove_duplicate_elements_from_vector(split_points);
     split_points.shrink_to_fit();
 
     MatrixXd X{MatrixXd::Constant(split_points.size(), number_of_base_terms, 0)};

cpp/constants.h

@@ -5,4 +5,5 @@ const double NAN_DOUBLE{std::numeric_limits<double>::quiet_NaN()};
 const int MAX_ABS_EXPONENT_TO_APPLY_ON_LINEAR_PREDICTOR_IN_LOGIT_MODEL{std::min(16, std::numeric_limits<double>::max_exponent10)};
 const std::string MSE_LOSS_FUNCTION{"mse"};
 const size_t MIN_CATEGORIES_IN_CLASSIFIER{2};
+const double DIVISOR_IN_GET_MAIN_EFFECT_SHAPE_FUNCTION{1000.0};
 const Eigen::Index MIN_OBSERATIONS_IN_A_CV_FOLD{2};

cpp/tests.cpp

@@ -167,7 +167,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 18.534016846656947));
+        tests.push_back(is_approximately_equal(predictions.mean(), 19.067710451454566));
     }
 
     void test_aplrregressor_cauchy_penalties()
@@ -221,7 +221,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 20.146282076477394));
+        tests.push_back(is_approximately_equal(predictions.mean(), 20.809163574542939));
     }
 
     void test_aplrregressor_cauchy_linear_effects_only_first()
@@ -275,7 +275,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 17.964887018234787));
+        tests.push_back(is_approximately_equal(predictions.mean(), 17.965154984786622));
     }
 
     void test_aplrregressor_cauchy_group_mse_validation()
@@ -466,7 +466,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 20.873594934501561));
+        tests.push_back(is_approximately_equal(predictions.mean(), 20.979930894644177));
     }
 
     void test_aplrregressor_custom_loss_and_validation()
@@ -526,7 +526,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.91507568241019));
+        tests.push_back(is_approximately_equal(predictions.mean(), 23.87336747209412));
     }
 
     void test_aplrregressor_custom_loss()
@@ -585,7 +585,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.703500296203778, 0.00001));
+        tests.push_back(is_approximately_equal(predictions.mean(), 24.301339246925711, 0.00001));
     }
 
     void test_aplrregressor_gamma_custom_link()
@@ -913,7 +913,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 20.82771158964184));
+        tests.push_back(is_approximately_equal(predictions.mean(), 20.849747430496922));
     }
 
     void test_aplrregressor_group_mse_cycle()
@@ -957,7 +957,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.526475166355244));
+        tests.push_back(is_approximately_equal(predictions.mean(), 23.529085584946195));
     }
 
     void test_aplrregressor_int_constr()
@@ -1010,7 +1010,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.576830262038001));
+        tests.push_back(is_approximately_equal(predictions.mean(), 23.657546542794449));
     }
 
     void test_aplrregressor_inversegaussian()
@@ -1171,7 +1171,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.563270291507191));
+        tests.push_back(is_approximately_equal(predictions.mean(), 23.602543167509292));
     }
 
     void test_aplrregressor_monotonic()
@@ -1224,7 +1224,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.47597042545404));
+        tests.push_back(is_approximately_equal(predictions.mean(), 23.34283475003015));
     }
 
     void test_aplrregressor_monotonic_ignore_interactions()
@@ -1492,7 +1492,7 @@ class Tests
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.610872525541577));
+        tests.push_back(is_approximately_equal(predictions.mean(), 23.646255799722155));
     }
 
     void test_aplrregressor_weibull()
@@ -1597,19 +1597,19 @@ class Tests
 
         VectorXd predictions{model.predict(X_test)};
         MatrixXd li{model.calculate_local_feature_contribution(X_test)};
-        VectorXd li_for_particular_terms{model.calculate_local_contribution_from_selected_terms(X_train, {5, 1})};
+        VectorXd li_for_particular_terms{model.calculate_local_contribution_from_selected_terms(X_train, {1, 8})};
 
         // Saving results
         save_as_csv_file("data/output.csv", predictions);
 
         std::cout << predictions.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(predictions.mean(), 23.703500296203778, 0.00001));
+        tests.push_back(is_approximately_equal(predictions.mean(), 24.301339246925711, 0.00001));
 
         std::map<double, double> main_effect_shape = model.get_main_effect_shape(1);
-        bool main_effect_shape_has_correct_length{main_effect_shape.size() == 11};
-        bool main_effect_shape_value_test{is_approximately_equal(main_effect_shape.begin()->second, -0.44924570143235887)};
+        bool main_effect_shape_has_correct_length{main_effect_shape.size() == 9};
+        bool main_effect_shape_value_test{is_approximately_equal(main_effect_shape.begin()->second, 0)};
         bool li_for_particular_terms_has_correct_size{li_for_particular_terms.rows() == X_train.rows()};
-        bool li_for_particular_terms_mean_is_correct{is_approximately_equal(li_for_particular_terms.mean(), 0.30321952178814915)};
+        bool li_for_particular_terms_mean_is_correct{is_approximately_equal(li_for_particular_terms.mean(), -0.52786383485971788)};
         tests.push_back(main_effect_shape_has_correct_length);
         tests.push_back(main_effect_shape_value_test);
         tests.push_back(li_for_particular_terms_has_correct_size);
@@ -2023,15 +2023,15 @@ class Tests
 
         std::cout << "cv_error\n"
                   << model.get_cv_error() << "\n\n";
-        tests.push_back(is_approximately_equal(model.get_cv_error(), 0.16317052975361318, 0.000001));
+        tests.push_back(is_approximately_equal(model.get_cv_error(), 0.15984656957508173, 0.000001));
 
         std::cout << "predicted_class_prob_mean\n"
                   << predicted_class_probabilities.mean() << "\n\n";
         tests.push_back(is_approximately_equal(predicted_class_probabilities.mean(), 0.5, 0.00001));
 
         std::cout << "local_feature_importance_mean\n"
                   << local_feature_importance.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(local_feature_importance.mean(), 0.054997728196581296, 0.00001));
+        tests.push_back(is_approximately_equal(local_feature_importance.mean(), 0.052181259967961045, 0.00001));
     }
 
     void test_aplrclassifier_two_class_predictor_specific_penalties_and_learning_rates()
@@ -2093,15 +2093,15 @@ class Tests
 
         std::cout << "cv_error\n"
                   << model.get_cv_error() << "\n\n";
-        tests.push_back(is_approximately_equal(model.get_cv_error(), 0.16984042158451909, 0.000001));
+        tests.push_back(is_approximately_equal(model.get_cv_error(), 0.17250319103503037, 0.000001));
 
         std::cout << "predicted_class_prob_mean\n"
                   << predicted_class_probabilities.mean() << "\n\n";
         tests.push_back(is_approximately_equal(predicted_class_probabilities.mean(), 0.5, 0.00001));
 
         std::cout << "local_feature_importance_mean\n"
                   << local_feature_importance.mean() << "\n\n";
-        tests.push_back(is_approximately_equal(local_feature_importance.mean(), 0.076147629914484025, 0.00001));
+        tests.push_back(is_approximately_equal(local_feature_importance.mean(), 0.07920242388299352, 0.00001));
     }
 
     void test_aplrclassifier_two_class_max_terms()

documentation/APLR 10.2.0.pdf renamed to documentation/APLR 10.2.1.pdf

@@ -25,7 +25,7 @@
 
 setuptools.setup(
     name="aplr",
-    version="10.2.0",
+    version="10.2.1",
     description="Automatic Piecewise Linear Regression",
     ext_modules=[sfc_module],
     author="Mathias von Ottenbreit",